CVPR2016跟踪算法Staple的配置（Staple: Complementary Learners for Real-Time Tracking）

代码下载地址：

https://github.com/bertinetto/staple

由于Staple运行需要特定的数据格式，本文把原工程目录runTracker.m简单修改便于测试，改动不大。

1、运行测试自带视频例子

在工程目录下新建run_example1.m文件

function run_example1
% RUN_TRACKER  is the external function of the tracker - does initialization and calls trackerMain%% Read params.txtparams = readParams('params.txt');%% load video infostart_frame=1;                      %开始帧sequence_path = ['vot15_ball1/'];   %视频路径总路径img_path = [sequence_path 'imgs/']; %图片路径%% Read filestext_files = dir([sequence_path '*_frames.txt']);f = fopen([sequence_path text_files(1).name]);frames = textscan(f, '%f,%f');    %一共多少帧if exist('start_frame')frames{1} = start_frame;elseframes{1} = 1;endfclose(f);params.bb_VOT = csvread([sequence_path 'groundtruth.txt']);region = params.bb_VOT(frames{1},:);%%%%%%%%%%%%%%%%%%%%%%%%%% read all the frames in the 'imgs' subfolderdir_content = dir([sequence_path 'imgs/']);% skip '.' and '..' from the countn_imgs = length(dir_content) - 2;img_files = cell(n_imgs, 1);for ii = 1:n_imgsimg_files{ii} = dir_content(ii+2).name;endimg_files(1:start_frame-1)=[];im = imread([img_path img_files{1}]);% is a grayscale sequence ?if(size(im,3)==1)params.grayscale_sequence = true;endparams.img_files = img_files;params.img_path = img_path;%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%if(numel(region)==8)% polygon format[cx, cy, w, h] = getAxisAlignedBB(region);elsex = region(1);y = region(2);w = region(3);h = region(4);cx = x+w/2;cy = y+h/2;end% init_pos is the centre of the initial bounding boxparams.init_pos = [cy cx];params.target_sz = round([h w]);[params, bg_area, fg_area, area_resize_factor] = initializeAllAreas(im, params);if params.visualizationparams.videoPlayer = vision.VideoPlayer('Position', [100 100 [size(im,2), size(im,1)]+30]);end% in runTracker we do not output anythingparams.fout = -1;% start the actual trackingtrackerMain(params, im, bg_area, fg_area, area_resize_factor);fclose('all');
end

2、运行run_example1.m

3、跑OTB数据集

在工程目录下新建run_example2.m文件

function run_example2
% RUN_TRACKER  is the external function of the tracker - does initialization and calls trackerMain%% Read params.txtparams = readParams('params.txt');%% load video infovideos = {'David','Basketball', 'Bolt'};start_frame=1;
%     videos = {'Basketball', 'Bolt', 'Boy', 'Car4', 'CarDark', 'CarScale', ...
%     'Coke', 'Couple', 'Crossing', 'David2', 'David3', 'David', 'Deer', ...
%     'Dog1', 'Doll', 'Dudek', 'Faceocc1', 'Faceocc2', 'Fish', 'Fleetface', ...
%     'Football', 'Football1', 'Freeman1', 'Freeman3', 'Freeman4', 'Girl', ...
%     'Ironman', 'Jogging_1', 'Jumping', 'Lemming', 'Liquor', 'Matrix', ...
%     'Mhyang', 'MotorRolling', 'MountainBike', 'Shaking', 'Singer1', ...
%     'Singer2', 'Skating1', 'Skiing', 'Soccer', 'Subway', 'Suv', 'Sylvester', ...
%     'Tiger1', 'Tiger2', 'Trellis', 'Walking', 'Walking2', 'Woman'};
for s=1:numel(videos)sequence_path = ['D:/Datasets/OTB100/',videos{s},'/'];   %视频路径总路径img_path = [sequence_path 'img/']; %图片路径params.bb_VOT = csvread([sequence_path 'groundtruth_rect.txt']);region = params.bb_VOT(start_frame,:);%%%%%%%%%%%%%%%%%%%%%%%%%% read all the frames in the 'imgs' subfolderdir_content = dir([sequence_path 'img/']);% skip '.' and '..' from the countn_imgs = length(dir_content) - 2;img_files = cell(n_imgs, 1);for ii = 1:n_imgsimg_files{ii} = dir_content(ii+2).name;endimg_files(1:start_frame-1)=[];im = imread([img_path img_files{1}]);% is a grayscale sequence ?if(size(im,3)==1)params.grayscale_sequence = true;endparams.img_files = img_files;params.img_path = img_path;%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%if(numel(region)==8)% polygon format[cx, cy, w, h] = getAxisAlignedBB(region);elsex = region(1);y = region(2);w = region(3);h = region(4);cx = x+w/2;cy = y+h/2;end% init_pos is the centre of the initial bounding boxparams.init_pos = [cy cx];params.target_sz = round([h w]);[params, bg_area, fg_area, area_resize_factor] = initializeAllAreas(im, params);if params.visualizationparams.videoPlayer = vision.VideoPlayer('Position', [100 100 [size(im,2), size(im,1)]+30]);end% in runTracker we do not output anythingparams.fout = -1;% start the actual trackingtrackerMain(params, im, bg_area, fg_area, area_resize_factor);fclose('all');
end
end

更改第七行为你想测试的OTB视频序列名称

更改第十八行为数据集所在路径

4、运行run_example2.m

5、跑VOT数据集

在工程目录下新建run_example3.m文件

function run_example3
% RUN_TRACKER  is the external function of the tracker - does initialization and calls trackerMain%% Read params.txtparams = readParams('params.txt');%% load video infovideos = {'bag','ball1','ball2'};start_frame=1;
%     videos = {'bag','ball1','ball2','basketball','birds1','birds2','blanket','bmx','bolt1',...
%               'bolt2','book','butterfly','car1','car2','crossing','dinosaur','fernando','fish1',...
%               'fish2','fish3','fish4','girl','glove','godfather','graduate','gymnastics1',...
%               'gymnastics2','gymnastics3','gymnastics4','hand','handball1','handball2',...
%               'helicopter','iceskater1','iceskater2','leaves','marching','matrix','motocross1',...
%               'motocross2','nature','octopus','pedestrian1','pedestrian2','rabbit','racing',...
%               'road','shaking','sheep','singer1','singer2','singer3','soccer1','soccer2',...
%               'soldier','sphere','tiger','traffic','tunnel','wiper'};
for s=1:numel(videos)sequence_path = ['D:/Datasets/vot2016/',videos{s},'/'];   %视频路径总路径img_path = sequence_path; %图片路径params.bb_VOT = csvread([img_path 'groundtruth.txt']);region = params.bb_VOT(start_frame,:);%%%%%%%%%%%%%%%%%%%%%%%%%% read all the frames in the 'imgs' subfolderdir_content = dir(sequence_path);% skip '.' and '..' from the countn_imgs = length(dir_content) - 2;img_files = cell(n_imgs, 1);for ii = 1:n_imgsimg_files{ii} = dir_content(ii+2).name;endimg_files(1:start_frame-1)=[];im = imread([img_path img_files{1}]);% is a grayscale sequence ?if(size(im,3)==1)params.grayscale_sequence = true;endparams.img_files = img_files;params.img_path = img_path;%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%if(numel(region)==8)% polygon format[cx, cy, w, h] = getAxisAlignedBB(region);elsex = region(1);y = region(2);w = region(3);h = region(4);cx = x+w/2;cy = y+h/2;end% init_pos is the centre of the initial bounding boxparams.init_pos = [cy cx];params.target_sz = round([h w]);[params, bg_area, fg_area, area_resize_factor] = initializeAllAreas(im, params);if params.visualizationparams.videoPlayer = vision.VideoPlayer('Position', [100 100 [size(im,2), size(im,1)]+30]);end% in runTracker we do not output anythingparams.fout = -1;% start the actual trackingtrackerMain(params, im, bg_area, fg_area, area_resize_factor);fclose('all');
end
end

更改第七行为你想测试的VOT视频序列名称

更改第十八行为数据集所在路径

6、运行run_example3.m

CVPR2016跟踪算法Staple的配置（Staple: Complementary Learners for Real-Time Tracking）相关推荐

PAMI2018跟踪算法HCFTstar的配置（Robust Visual Tracking via Hierarchical Convolutional Features）
文章和代码下载地址: HCFTstar: Chao Ma, Jia-Bin Huang, Xiaokang Yang, Ming-Hsuan Yang. "Robust Visual Tra ...
CVPR2021跟踪算法TransT的配置（Transformer Tracking）
1.论文下载地址 TransT: Transformer Tracking.[paper] 2.代码下载地址 https://github.com/chenxin-dlut/TransT 3.建立虚拟 ...
CVPR2018跟踪算法VITAL的配置（VITAL: VIsual Tracking via Adversarial Learning）
工程介绍地址: VITAL 代码下载地址: https://github.com/ybsong00/Vital_release 一.跟踪测试 1.运行compile_matconvnet.m编译mat ...
PAMI2012跟踪算法TLD的配置（Tracking-learning-detection）
代码下载地址: https://github.com/zk00006/OpenTLD 1.打开compile.m更改OpenCV路径 2.运行compile.m 可能遇到错误: 原因:OpenCV版本 ...
CVPR2021跟踪算法STMTrack的配置（Template-free Visual Tracking with Space-time Memory Networks）
1.论文下载地址 STMTrack: Template-free Visual Tracking with Space-time Memory Networks CVPR (2021). [paper ...
OTB数据集和VOT数据集融合跟踪算法接口示例
OTB数据集和VOT数据集自己存数据接口参考代码: 一.OTB数据集(不使用tracker_benchmark_v1.0) 1.OTB数据结果最基本的格式 type为目标框类型: res为目标框的所在 ...
目标跟踪算法--Staple: Complementary Learners for Real-Time Tracking
文章下载链接:文章下载链接代码下载链接:Staple代码 -------------------------------------------- 今天要讲的这篇文章也是基于相关滤波器(不懂相关滤波 ...
Staple 跟踪: Complementary Learners for Real-Time Tracking
目标跟踪算法--Staple: Complementary Learners for Real-Time Tracking 小小菜鸟一只 2017-03-25 09:26:42 15110 收藏 ...
STAPLE目标跟踪算法(C++ /OpenCV)
简介结合梯度特征HOG及颜色特征的实时跟踪算法,速度可以达到80FPS. 摘要近些年来,基于相关滤波跟踪器的算法得到了很大的改进,已经能够达到很好的跟踪效果,14年的KCF,15年的SRDCF.H ...

CVPR2016跟踪算法Staple的配置（Staple: Complementary Learners for Real-Time Tracking）

CVPR2016跟踪算法Staple的配置（Staple: Complementary Learners for Real-Time Tracking）相关推荐

最新文章

热门文章